xAI has rolled out Grok Imagine Image 2.0 as Grok’s new Quality Mode, putting greater emphasis on precise edits, typography, layout consistency and production-ready creative workflows.
xAI has made Imagine Image 2.0 generally available, expanding Grok’s image-generation tools with a set of features aimed less at one-off AI pictures and more at the repeated editing that goes into finished creative work.
The new model is now available as Quality Mode on grok.com/imagine and through Grok’s iOS and Android apps. xAI describes the release as an attempt to make generated images more practical for real creative projects, particularly where users need to preserve details while making multiple changes to the same asset.
That distinction sits at the centre of Image 2.0. Rather than focusing only on generating a convincing first image, xAI is putting considerable weight behind what happens afterwards: changing a specific part of an image without disturbing everything around it.
Editing becomes a bigger part of Grok Imagine
One of the more notable additions is multi-reference editing, which can accept up to five input images in a single generation. The feature is designed to reduce the need to manually combine several assets before bringing them into an image-generation workflow.
Image 2.0 also includes a magic wand for editing a selected region while leaving the rest of an image intact, along with segmentation tools that allow users to identify precise areas they want to modify.
A background-removal option can isolate a subject and export it against a transparent background, making the resulting asset easier to place inside another design or composition.
Taken together, those tools show where xAI is trying to move Imagine: from a prompt-and-result experience toward something closer to an iterative visual workspace.
Smart Resize extends an image instead of simply cropping it
Image 2.0 also introduces Smart Resize, which allows a user to choose a different aspect ratio and have the model fill the additional frame.
The interface shown by xAI includes nine ratios: 1:2, 9:16, 2:3, 3:4, 1:1, 4:3, 3:2, 16:9 and 2:1. Rather than treating one generated image as locked to its original canvas, the feature is intended to make the same visual adaptable for different formats.
That could be particularly useful for work that begins with one creative direction but eventually needs several versions for different screen sizes or layouts.
xAI puts typography and layout high on the agenda
Another focus of the release is instruction-following.
xAI says Image 2.0 was built to follow prompts closely enough to preserve small details while handling more complicated compositions. The company also highlights typography and layout planning, saying the model is designed to keep dense, multi-part visuals coherent and produce sharper small text.
It is a meaningful area of focus because Image 2.0 is being positioned not only for photography-style outputs, but also for design and illustration. Editing was treated as a core capability during training rather than being added solely as a secondary feature, according to xAI.
xAI cites second-place Arena rankings
Alongside the product release, xAI says Image 2.0 ranks second globally in both text-to-image generation and image editing, citing Arena Image Edit and Text-to-Image leaderboards dated August 7, 2026.
The leaderboard graphic published with the announcement shows grok-imagine-image-2 (low) with an overall Elo score of 1,439 in Image Edit Arena, behind OpenAI’s gpt-image-2 at 1,463.
On the Text-to-Image Arena chart, grok-imagine-image-2 (low) is shown at 1,320, again behind gpt-image-2 at 1,380. xAI notes that its models appear on Arena under the SpaceXAI name.
The rankings are presented by xAI as evidence of Image 2.0’s performance, although the company’s announcement does not provide a separate independent evaluation beyond the cited Arena leaderboards.
Templates turn common creative jobs into starting points
The release also adds templates built around recurring image tasks.
Instead of starting each project from a blank prompt, users can choose workflows configured for jobs including photo editing, product shots, headshots, icons and game assets, then provide the required inputs.
xAI says the idea is to package common image workflows into ready-made starting points that can produce a finished result with less manual setup.
The examples shown with the launch range from design-oriented tools such as photo collages and mascot creation to broader visual-development work.
Image 2.0 also targets consistent worlds for video projects
One of the more interesting demonstrations goes beyond individual images.
xAI shows a workflow in which a character, locations and props are generated separately while maintaining a consistent visual style from image to image. The company presents this as a way to build a coherent visual world from a relatively small collection of prompts.
The accompanying examples include the same character across different snowy environments, together with separate location and prop images. It illustrates how consistency across generations has become part of the product’s pitch, rather than simply the quality of any single output.
Available now for consumers, API still to come
Imagine Image 2.0 can currently be used through Grok.com, Grok on iOS and Grok on Android.
Developers will have to wait longer. xAI says API access is coming soon, but the announcement does not provide a release date for it.
For now, the release is clearly centred on consumer-facing creative tools. Its mix of targeted editing, multi-image input, resizing, templates and cross-image consistency suggests that xAI wants Imagine to compete not just on how well an image begins, but on how usable that image remains once real editing starts.
Also Read: Jorge Messi, Father of Lionel Messi, Dies at 68 in Argentina
